Methods are being investigated to enable the NASArnEngineering and Safety Center (NESC) to becomernproactive in identifying unknown indicators of futurernproblems. Electronic data mining is a particularlyrnpromising tool for this effort into unsupervised learningrnof common factors. This work in progress began withrna systematic evaluation of available data miningrnsoftware packages and continues with in-depthrnbenchmarking of chosen candidates. Preliminaryrnrecommendations for best practices in data mining andrntrending are provided. This critical first step inrnidentifying unknown unknowns before they becomernproblems is applicable to any set of engineering orrnprogrammatic data.
展开▼
